王行漢,叢沛桐,亢 慶,扶卿華,劉超群,王曉剛
(1. 華南農(nóng)業(yè)大學(xué)水利與土木工程學(xué)院,廣州 510610; 2. 珠江水利委員會(huì)珠江水利科學(xué)研究院,廣州 510611;3. 水利部珠江河口動(dòng)力學(xué)及伴生過(guò)程調(diào)控重點(diǎn)實(shí)驗(yàn)室,廣州 510611)
·爭(zhēng)鳴與討論·
非線性擬合LST/NDVI特征空間干濕邊優(yōu)于傳統(tǒng)線性擬合方法的討論
王行漢1,2,3,叢沛桐1※,亢 慶2,扶卿華2,劉超群2,王曉剛2
(1. 華南農(nóng)業(yè)大學(xué)水利與土木工程學(xué)院,廣州 510610; 2. 珠江水利委員會(huì)珠江水利科學(xué)研究院,廣州 510611;3. 水利部珠江河口動(dòng)力學(xué)及伴生過(guò)程調(diào)控重點(diǎn)實(shí)驗(yàn)室,廣州 510611)
地表溫度/植被指數(shù)特征空間在土壤含水率、蒸散發(fā)等定量遙感反演和旱情監(jiān)測(cè)、水資源管理方面有著重要的應(yīng)用,但其特征空間中干濕邊的擬合方式的研究目前還相對(duì)缺乏。該文以美國(guó)俄克拉荷馬州為例,針對(duì)地表溫度/植被指數(shù)特征空間干邊和濕邊的最優(yōu)擬合方式展開研究,分別采用線性、指數(shù)、對(duì)數(shù)、多項(xiàng)式和冪函數(shù)對(duì)干邊和濕邊進(jìn)行擬合,并采用16個(gè)土壤墑情站點(diǎn)的5、25和60 cm不同深度的3組實(shí)測(cè)土壤含水率數(shù)據(jù)對(duì)擬合結(jié)果進(jìn)行評(píng)估。結(jié)果表明:對(duì)于干邊的擬合,指數(shù)函數(shù)、線性函數(shù)、對(duì)數(shù)函數(shù)和冪函數(shù)擬合的決定系數(shù)r2分別為0.64,0.60,0.41,0.43,多項(xiàng)式函數(shù)擬合的r2最高(0.67);對(duì)于濕邊的擬合,指數(shù)函數(shù)、線性函數(shù)、對(duì)數(shù)函數(shù)和冪函數(shù)擬合的r2分別為0.59,0.63,0.67,0.69,多項(xiàng)式函數(shù)擬合的r2最高,為0.70;多項(xiàng)式函數(shù)擬合干邊和濕邊構(gòu)建特征空間計(jì)算結(jié)果的均方根誤差(RMSE,root mean square error)和平均絕對(duì)誤差(MAE,mean absolute error)值均最小,在5、25和60 cm深度下RMSE分別為0.29、0.27和0.28,MAE分別為0.26、0.23和0.25,表明采用多項(xiàng)式函數(shù)擬合干邊和濕邊計(jì)算的結(jié)果精度最高且對(duì)25cm深度的土壤含水率最為敏感。
植被;溫度;土壤水分;地表溫度/植被指數(shù);特征空間;干邊;濕邊
王行漢,叢沛桐,亢 慶,扶卿華,劉超群,王曉剛.非線性擬合 LST/NDVI特征空間干濕邊優(yōu)于傳統(tǒng)線性擬合方法的討論[J]. 農(nóng)業(yè)工程學(xué)報(bào),2017,33(11):306-314. doi:10.11975/j.issn.1002-6819.2017.11.039 http://www.tcsae.org
Wang Xinghan, Cong Peitong, Kang Qing, Fu Qinghua, Liu Chaoqun, Wang Xiaogang. Discussion on method of nonlinear fitting dry and wet edges of LST/ NDVI feature space better than traditional linear fitting method [J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2017, 33(11): 306-314. (in Chinese with English abstract) doi:10.11975/j.issn.1002-6819.2017.11.039 http://www.tcsae.org
地表溫度/植被指數(shù)特征空間是對(duì)象元的地表輻射溫度和植被覆蓋度的二維空間分布的一種解譯[1]。三角形特征空間的概念是Price在研究區(qū)域尺度的蒸散發(fā)的過(guò)程中首次提出的[1-2],當(dāng)區(qū)域內(nèi)植被覆蓋和土壤干濕程度跨度范圍較大時(shí),遙感植被指數(shù)(vegetation index, VI)和地表溫度(land surface temperature, LST)的散點(diǎn)圖近似呈三角形分布,它包括2個(gè)極端的邊界,一個(gè)是干邊,代表水分嚴(yán)重虧缺,無(wú)水分可以用于蒸散;另一個(gè)是濕邊,代表土壤水分充足,植被處于潛在蒸散狀態(tài)。在早期的研究中,研究者發(fā)現(xiàn)遙感光譜植被指數(shù)和地表輻射溫度之間有著很強(qiáng)的相關(guān)性[3-4],借此用來(lái)對(duì)地表蒸散發(fā)和土壤含水率的研究,并進(jìn)一步的提出了以地表溫度/植被指數(shù)為代表的特征空間。特征空間方法依賴的基礎(chǔ)理論為:1)遙感植被指數(shù)和地表溫度之間有著強(qiáng)相關(guān)性[3-6];2)在同等植被覆蓋條件下,當(dāng)作物受到水分脅迫時(shí),葉片氣孔自衛(wèi)性關(guān)閉,減少了植被蒸騰作用,引起地表溫度上升。不少學(xué)者對(duì)于地表溫度/植被指數(shù)之間的關(guān)系進(jìn)行了研究[7-13],結(jié)果均表明其之間具有極強(qiáng)的負(fù)相關(guān)性[3-6],并進(jìn)一步的對(duì)地表溫度/植被指數(shù)與空氣氣溫[14]、土壤含水率[11,15]等之間的關(guān)系進(jìn)行了研究,被廣泛應(yīng)用于蒸散發(fā)[16-17]、土壤含水率[11]、土壤鹽漬化[18]、農(nóng)作物生長(zhǎng)[19]等參數(shù)的估算,對(duì)研究農(nóng)業(yè)灌溉、干旱監(jiān)測(cè)[20-23]、土地覆蓋利用變化監(jiān)測(cè)以及水資源管理與規(guī)劃具有重要的意義。
利用特征空間估算蒸散發(fā)和土壤含水率的關(guān)鍵在于干邊和濕邊的確定,也就是獲取充分供水及極端缺水條件下植被指數(shù)與地表溫度的計(jì)算關(guān)系。過(guò)往的研究大體可以分為2種思路,一種是依賴于經(jīng)驗(yàn)和先驗(yàn)知識(shí),在干邊上選擇地表溫度最大值[24-25],在濕邊上主要選擇水體表面溫度[26]、區(qū)域平均氣溫[27]或者地表溫度最小值[24-25]。另一種思路是通過(guò)一定原理計(jì)算特征空間頂點(diǎn)的地表溫度或者地表溫度與大氣溫差[27-29]。對(duì)于特征空間的構(gòu)建,其干邊和濕邊的確定尤為關(guān)鍵,但是目前對(duì)于干邊和濕邊擬合方式方面的研究還相對(duì)缺乏。在傳統(tǒng)的地表溫度/植被指數(shù)特征空間模型的構(gòu)建當(dāng)中,對(duì)于模型中的干邊基本均采取了簡(jiǎn)單的線性負(fù)相關(guān)關(guān)系進(jìn)行參數(shù)擬合,對(duì)于濕邊認(rèn)為與坐標(biāo)軸平行的簡(jiǎn)化處理[11]。這樣的處理是否恰當(dāng),精度如何是本文著重需要進(jìn)行討論的。
本文針對(duì)地表溫度/植被指數(shù)特征空間干邊和濕邊的擬合方式展開研究,分別采用線性、指數(shù)、對(duì)數(shù)、多項(xiàng)式和冪函數(shù)對(duì)干邊和濕邊分別進(jìn)行擬合,并結(jié)合美國(guó)西南部地區(qū)俄克拉荷馬州的一塊研究區(qū),利用實(shí)測(cè)土壤含水率數(shù)據(jù)對(duì)擬合結(jié)果進(jìn)行評(píng)估,以探討和提高特征空間方法在大空間尺度上的適用性。
1.1 研究區(qū)
本研究中選取美國(guó)俄克拉荷馬州一塊158 km× 147 km區(qū)域?yàn)檠芯繉?duì)象,其中心位置為36°2′12.48″N、97°49′42.71″W,如圖1所示。該區(qū)域內(nèi)以平原為主,海拔在252~592 m之間。氣候以干燥的溫帶大陸性氣候?yàn)橹?,年平均氣溫?5.5 ℃,降雨量自東向西遞減。森林覆蓋率約為24%,主要分布在區(qū)域東部,研究區(qū)內(nèi)的中北部主要為農(nóng)業(yè)用地,植被覆蓋度相對(duì)較低。
圖1 研究區(qū)及土壤墑情站點(diǎn)分布圖Fig.1 Location of study area and distribution of soil moisture stations
1.2 數(shù)據(jù)源
1)遙感數(shù)據(jù)
通過(guò)美國(guó)地質(zhì)調(diào)查局(USGS)網(wǎng)站(http://glovis.usgs.gov/),下載獲取研究區(qū)內(nèi)Landsat TM5衛(wèi)星數(shù)據(jù),成像時(shí)間為2009年9月28日,影像質(zhì)量為最佳等級(jí),無(wú)云。Landsat TM5包含可見光-熱紅外多個(gè)光譜通道和熱紅外數(shù)據(jù),空間分辨率為30 m,詳細(xì)參數(shù)見表1。
表1 Landsat TM5遙感影像信息Table 1 Information of Landsat TM5 image
2)土壤墑情數(shù)據(jù)
土壤墑情數(shù)據(jù)來(lái)源于美國(guó)National integrated drought information system(https://www.drought.gov/drought/ soil-moisture-map),該系統(tǒng)包含了一系列深度的土壤含水率數(shù)據(jù)。在該研究區(qū)域內(nèi)設(shè)有16個(gè)土壤墑情觀測(cè)站(詳細(xì)信息見表2),各站點(diǎn)的空間位置見圖1。本文中使用5、25和60 cm共3種不同深度的土壤含水率數(shù)據(jù)對(duì)特征空間計(jì)算的結(jié)果進(jìn)行分析。
表2 土壤墑情站點(diǎn)信息Table 2 Information of soil moisture stations
3)土壤類型數(shù)據(jù)
土壤類型數(shù)據(jù)來(lái)源于聯(lián)合國(guó)糧農(nóng)組織(FAO)和維也納國(guó)際應(yīng)用系統(tǒng)研究所(IIASA)所構(gòu)建的世界土壤數(shù)據(jù)庫(kù)(Harmonized World Soil Database version 1.2)(http://webarchive.iiasa.ac.at/Research/LUC/External-Worldsoil-database)。通過(guò)查詢,在該研究區(qū)內(nèi)土壤類型單一,均為栗鈣土(Kastanozems)。該土壤類型富含腐殖質(zhì),最初覆蓋著早熟的原生草原植被,其深度在第1米內(nèi)產(chǎn)生特征為棕色的表面層,具有與土壤顆粒結(jié)合相對(duì)高水平的可用鈣離子,并且可以具有25~100 cm之間厚度的石炭蝕層。
1.3 歸一化植被指數(shù)的計(jì)算
在本研究中,特征空間橫坐標(biāo)的植被指數(shù)的選取采用的是目前國(guó)際上使用最為廣泛的植被指數(shù)——?dú)w一化植被指數(shù)(normailzed difference vegetation index, NDVI),它對(duì)于植被的生長(zhǎng)和空間分布具有很好的指示作用,與植被覆蓋度有很強(qiáng)的相關(guān)性[30-31]。其數(shù)學(xué)表達(dá)式為
式中ρNIR為近紅外波段,ρRED為紅光波段。NDVI理論值處于[-1,1]之間,負(fù)值表示地面覆蓋為云、水、雪等類型,對(duì)可見光高反射;零值表示有巖石或裸土等類型,近紅外波段光譜值和紅波段光譜值近似相等;正值表示有植被覆蓋,且隨覆蓋度增大而增大。但是由于受到地表濕度、太陽(yáng)光照條件和大氣條件等的影響,存在一些異常值,在本研究區(qū)域內(nèi)負(fù)值主要為水體,在構(gòu)建特征空間時(shí)對(duì)其進(jìn)行了掩膜處理。
1.4 地表溫度的計(jì)算
目前常用的地表溫度的計(jì)算方法包括輻射傳輸方程算法、單通道算法和分裂窗算法,前人針對(duì)這3種計(jì)算方法開展了一系列研究[32],結(jié)果表明輻射傳輸方程算法和分裂窗算法的精度相對(duì)較高,并且輻射傳輸方程算法的物理基礎(chǔ)明確,反演精度較高。因此在本研究中選取了輻射傳輸方程算法反演地表溫度,其原理是通過(guò)估計(jì)大氣對(duì)地表熱輻射的影響,并將大氣對(duì)地表熱輻射的影響從衛(wèi)星傳感器所觀測(cè)到的熱輻射總量中減去,得到地表熱輻射強(qiáng)度,最后將熱輻射強(qiáng)度轉(zhuǎn)化為地表溫度。地表溫度LST計(jì)算的數(shù)學(xué)表達(dá)式為
式中K1、K2為傳感器的定標(biāo)常數(shù),B(Ts)指溫度為Ts的黑體輻射亮度,B(Ts)計(jì)算的數(shù)學(xué)表達(dá)式為
式中Lλ為熱紅外輻射亮度值,ε為地表比輻射率,τ為大氣在熱紅外波段的透過(guò)率,L↑大氣向上輻射亮度,L↓大氣向下輻射亮度。其中,大氣在熱紅外波段的透過(guò)率、大氣向上輻射亮度、大氣向下輻射亮度可以在NASA官網(wǎng)(http://atmcorr.gsfc.nasa.gov/)中輸入成影時(shí)間以及中心經(jīng)緯度獲得;地表比輻射率使用Sobrino等[33]提出的NDVI閾值法計(jì)算,數(shù)學(xué)表達(dá)式為
式中FVC是指植被覆蓋度。對(duì)于植被覆蓋度的計(jì)算采用的是基于NDVI的像元二分法模型[34-35],計(jì)算數(shù)學(xué)表達(dá)式為
式中NDVIsoil指的是裸地像元值,NDVIvegetation指的是植被覆蓋的像元值,目前常用的處理方式是根據(jù)整幅遙感影像上的NDVI的分布,以0.5%置信度[35]截取NDVI的上下閾值分別近似代表NDVIsoil和NDVIvegetation。
1.5 地表溫度/植被指數(shù)特征空間構(gòu)建及驗(yàn)證
地表溫度/植被指數(shù)特征空間構(gòu)建的核心是對(duì)干邊和濕邊的確定,傳統(tǒng)的方法采用的是線性擬合,也就是用固定的NDVI間隔切割NDVI與LST的散點(diǎn)圖,分別找到間隔內(nèi)地表溫度的最大值(LSTmax)和最小值(LSTmin),分別線性擬合NDVI值,得到干邊和濕邊。
本研究中采用同樣的方法獲取地表溫度的最大值和最小值,但假定擬合方式為非線性關(guān)系(包括指數(shù)關(guān)系、線性關(guān)系、對(duì)數(shù)關(guān)系、多項(xiàng)式關(guān)系和冪函數(shù)關(guān)系,見圖2),非線性擬合的數(shù)學(xué)表達(dá)式如下
式中α和β代表干邊和濕邊擬合方程的擬合系數(shù),n代表多項(xiàng)式的次數(shù)。
采用基于不同函數(shù)擬合方式構(gòu)建的LST/NDVI特征空間,計(jì)算土壤含水率,數(shù)學(xué)表達(dá)式如下
LST/NDVI代表土壤含水率,為介于0~1之間的無(wú)量綱值,其值越接近1土壤含水率越低,越接近0其土壤含水率越高。
圖2 地表溫度/植被指數(shù)特征空間Fig.2 LST/NDVI feature space
為了評(píng)估不同函數(shù)擬合地表溫度/植被指數(shù)特征空間計(jì)算結(jié)果的精度,本研究中采用決定系數(shù)r2、均方根誤差(root mean square error, RMSE)和平均絕對(duì)誤差(mean absolute error, MAE)作為評(píng)估的指標(biāo)參數(shù)。均方根誤差和平均絕對(duì)誤差主要用來(lái)衡量觀測(cè)值和模型值之間的偏差,能夠很好地反映出模型計(jì)算的精確程度,其值越小,精確度越高[36]。數(shù)學(xué)表達(dá)式為
式中α,iΧ代表真實(shí)值,β,iΧ代表模型計(jì)算值,n代表真實(shí)值的個(gè)數(shù)。
式中P和Q分別為實(shí)測(cè)值和估算值;PM和QM則分別為實(shí)測(cè)值和估算值的平均值;下標(biāo)i代表n個(gè)實(shí)測(cè)值或者估算值中的第i個(gè)值。
2.1 NDVI與LST計(jì)算結(jié)果分析
NDVI和LST作為地表溫度/植被指數(shù)特征空間構(gòu)建的2個(gè)關(guān)鍵參數(shù)。其中,NDVI為特征空間的橫坐標(biāo),LST為特征空間的縱坐標(biāo)。在本研究中,利用Landsat TM5衛(wèi)星遙感影像數(shù)據(jù),經(jīng)過(guò)大氣校正、輻射定標(biāo)等數(shù)據(jù)預(yù)處理后,結(jié)合公式(1)計(jì)算得到歸一化植被指數(shù)NDVI,結(jié)合公式(2)-(5)計(jì)算得到地表溫度LST,并借助于ENVI軟件平臺(tái)對(duì)其進(jìn)行進(jìn)行了統(tǒng)計(jì),結(jié)果見圖3和圖4。
圖3 歸一化植被指數(shù)計(jì)算結(jié)果Fig.3 Calculation results of normalized differential vegetation index
由圖3a可知,研究區(qū)內(nèi)植被指數(shù)NDVI值空間分布存在較大差異,在中北部地區(qū)為農(nóng)業(yè)用地集中分布區(qū)域,NDVI值主要集中在0.15左右,并且高植被指數(shù)與低植被指數(shù)呈現(xiàn)鑲嵌式分布的特點(diǎn);在東部地區(qū)主要為林草地覆蓋區(qū)域,NDVI值主要集中在0.55左右。根據(jù)研究區(qū)內(nèi)NDVI值統(tǒng)計(jì)曲線(圖3b)可知,研究區(qū)內(nèi)的NDVI值主要集中0.1~0.8之間,極大值和極小值的數(shù)量相對(duì)較少;并且NDVI值的數(shù)量在0.1~0.16和0.32~0.55兩個(gè)區(qū)間內(nèi)呈現(xiàn)不斷增加的變化趨勢(shì),在0.16~0.32和0.55~0.9兩個(gè)區(qū)間內(nèi)呈現(xiàn)不斷減少的變化趨勢(shì)。上述結(jié)果表明了在整個(gè)研究區(qū)內(nèi)體現(xiàn)了地表覆蓋從裸土到稀疏植被再到茂密植被的變化過(guò)程,滿足特征空間構(gòu)建要求植被具有不同覆蓋程度的要求。
由圖4a可知,研究區(qū)東部地區(qū)LST值相對(duì)較低,主要集中在20~25 ℃之間,在中北部地區(qū)LST值相對(duì)較高,主要集中在25~35 ℃之間,并且25~30 ℃和30~35 ℃兩個(gè)區(qū)間的LST值呈現(xiàn)鑲嵌分布的特征。根據(jù)研究區(qū)內(nèi)LST值統(tǒng)計(jì)曲線(圖4b)可知,研究區(qū)內(nèi)的LST值主要集中在20~35 ℃之間,而小于20 ℃和大于35 ℃的LST值的數(shù)量相對(duì)較少,并且LST值的變化范圍較寬,能夠滿足特征空間的構(gòu)建需要足夠值域范圍的地表溫度變化區(qū)間的要求。
圖4 地表溫度計(jì)算結(jié)果Fig.4 Calculation results of land surface temperature
因此,計(jì)算得到的NDVI和LST能夠滿足構(gòu)建地表溫度/植被指數(shù)的要求,進(jìn)一步的基于上述2個(gè)參數(shù)進(jìn)行干邊和濕邊擬合。
2.2 干濕邊擬合及LST/NDVI特征空間構(gòu)建
基于ENVI軟件平臺(tái),通過(guò)空間疊加分析,計(jì)算和統(tǒng)計(jì)NDVI對(duì)應(yīng)地表溫度的最大值和最小值,結(jié)果如圖5所示。在干邊上,即由最大地表溫度組成的離散點(diǎn),NDVI與LST呈現(xiàn)非線性關(guān)系,并且在NDVI為0.3~0.7區(qū)間內(nèi)非線性特征關(guān)系表現(xiàn)尤為明顯。在濕邊上,即由最小地表溫度組成的離散點(diǎn),在NDVI介于0~0.8之間,LST隨著NDVI的增加而增加,但超過(guò)0.8之后,LST值基本穩(wěn)定,主要原因是在高植被覆蓋區(qū),當(dāng)NDVI達(dá)到一定的值后發(fā)生了飽和的現(xiàn)象,這與王行漢等[37]研究的結(jié)果相一致。
圖5 研究區(qū)內(nèi)地表溫度/植被指數(shù)特征空間構(gòu)建Fig.5 Construction of LST/NDVI feature space in study area
根據(jù)上述構(gòu)建的研究區(qū)內(nèi)的LST/NDVI特征空間(圖5),分別采用線性、多項(xiàng)式、對(duì)數(shù)、指數(shù)和冪5種函數(shù)方式對(duì)干邊和濕邊進(jìn)行擬合,決定系數(shù)計(jì)算采用公式(10),結(jié)果見表3。
表3中5種不同函數(shù)擬合方式的決定系數(shù)r2介于0.4~0.7之間。對(duì)于干邊,多項(xiàng)式函數(shù)的擬合效果最好,r2為0.67,其次是指數(shù)函數(shù)r2為0.64,對(duì)數(shù)函數(shù)和冪函數(shù)擬合效果相對(duì)較差,兩者r2均不超過(guò)0.5。對(duì)于濕邊,多項(xiàng)式擬合的效果最好,r2為0.7,冪函數(shù)次之,指數(shù)函數(shù)擬合效果最差。綜合干邊和濕邊的擬合方式來(lái)看,多項(xiàng)式的擬合效果最好。
表3 干邊和濕邊5種函數(shù)擬合方程及其決定系數(shù)Table 3 Fitting equations of dry edge and wet edge and their determination coefficient of five functions
2.3 特征空間計(jì)算及驗(yàn)證分析
根據(jù)表3中線性、多項(xiàng)式、對(duì)數(shù)、指數(shù)和冪5種函數(shù)干邊和濕邊的擬合方程和地表溫度/植被指數(shù)的計(jì)算方法(公式(6)、公式(7)),計(jì)算得到研究區(qū)內(nèi)LST/NDVI值(圖6),結(jié)果表明:
5種函數(shù)的擬合均能夠體現(xiàn)出不同土壤含水率的空間分布,從宏觀角度看整體空間分布規(guī)律存在一致性,土壤含水率較低的地區(qū)主要集中在研究區(qū)的中北部、中西部,土壤含水率相對(duì)較高的地區(qū)集中在研究區(qū)的東部地區(qū)。但對(duì)于具體計(jì)算結(jié)果值的表現(xiàn)上,5種函數(shù)又有所差別,指數(shù)函數(shù)的計(jì)算結(jié)果主要集中在0.5~0.6之間(圖6a);線性函數(shù)的計(jì)算結(jié)果主要集中在0.45~0.55之間,數(shù)據(jù)值的空間分布上與指數(shù)函數(shù)一致性較高(圖6b);對(duì)數(shù)函數(shù)的計(jì)算結(jié)果主要集中在0.55~0.6之間,在0.2~0.4區(qū)間內(nèi)數(shù)值的空間突變性較指數(shù)函數(shù)、線性函數(shù)和冪函數(shù)更加突出(圖6c);冪函數(shù)的計(jì)算結(jié)果主要集中在0.6左右,0.2~0.6區(qū)間內(nèi)數(shù)值變化呈現(xiàn)平緩上升趨勢(shì),無(wú)明顯突變(圖6d);多項(xiàng)式函數(shù)的計(jì)算結(jié)果在0.2~0.8區(qū)間內(nèi)的突變現(xiàn)象比較明顯,并且數(shù)據(jù)結(jié)果主要集中在0.4~0.6之間(圖6e)。由于研究區(qū)內(nèi)中北部絕大部分地區(qū)為農(nóng)業(yè)用地,高植被區(qū)與低植被區(qū)呈現(xiàn)鑲嵌分布的特點(diǎn),因此容易造成計(jì)算結(jié)果出現(xiàn)突變的現(xiàn)象,基于多項(xiàng)式函數(shù)的計(jì)算結(jié)果更加符合上述特征。
圖6 研究區(qū)內(nèi)5種函數(shù)擬合地表溫度/植被指數(shù)計(jì)算結(jié)果Fig.6 LST/NDVI calculation results of five function fitting equations in study area
進(jìn)一步為對(duì)比采用指數(shù)函數(shù)、線性函數(shù)、多項(xiàng)式函數(shù)、對(duì)數(shù)函數(shù)和冪函數(shù)分別擬合干邊和濕邊構(gòu)建特征空間計(jì)算的LST/NDVI結(jié)果(圖6),利用ArcGIS軟件平臺(tái)的空間分析模塊提取16個(gè)土壤墑情站點(diǎn)對(duì)應(yīng)的5種函數(shù)擬合干濕邊計(jì)算的結(jié)果,并將該結(jié)果和土壤墑情站點(diǎn)的實(shí)測(cè)數(shù)據(jù)進(jìn)行對(duì)比,并采用RMSE、MAE等指標(biāo)參數(shù)對(duì)計(jì)算結(jié)果進(jìn)行精度分析。
由于研究區(qū)內(nèi)的土壤類型為栗鈣土,土壤類型單一,因此可以不考慮不同土壤類型的影響。土壤墑情站點(diǎn)觀測(cè)值為體積含水量,模型計(jì)算結(jié)果為相對(duì)含水量,在不考慮土壤質(zhì)地等因素的影響,兩者存在正比例關(guān)系,研究表明其之間存在一定相關(guān)性,因此可以作為精度評(píng)價(jià)的參考[11,20,22,38]。本研究中獲取的16個(gè)土壤墑情站點(diǎn)的觀測(cè)數(shù)據(jù)包含了5、25和60 cm 3種不同深度的土壤含水率值,通過(guò)利用土壤墑情站點(diǎn)不同深度的3組數(shù)據(jù)分別與5種函數(shù)擬合計(jì)算的結(jié)果進(jìn)行對(duì)比,計(jì)算得到RMSE和MAE,結(jié)果見表4。
表4 不同深度土壤的5種函數(shù)擬合方式的均方根誤差和平均絕對(duì)誤差Table 4 Root mean square error and mean absolute error of five function fitting equations of different soil layers
根據(jù)表4可知,5種函數(shù)擬合的干邊和濕邊計(jì)算特征空間的結(jié)果精度存在一定的差異。由RMSE的結(jié)果可以看出:與5 cm深度的土壤含水率進(jìn)行對(duì)比,多項(xiàng)式函數(shù)的RMSE值最小為0.29,其次是線性函數(shù)為0.31,對(duì)數(shù)函數(shù)為0.32,指數(shù)函數(shù)和冪函數(shù)最大;與25 cm深度的土壤含水率進(jìn)行對(duì)比,多項(xiàng)式函數(shù)的RMSE值最小為0.27,其次是線性函數(shù)和對(duì)數(shù)函數(shù),均為0.29,指數(shù)函數(shù)為0.30,冪函數(shù)最小為0.31;與60 cm深度的土壤含水率進(jìn)行對(duì)比,多項(xiàng)式函數(shù)的RMSE值最小為0.28,其次是線性函數(shù)為0.30,對(duì)數(shù)函數(shù)為0.31,冪函數(shù)和指數(shù)函數(shù)最小,均為0.32。由MAE的結(jié)果可以看出:5種函數(shù)中,對(duì)于5、25和60 cm 3種不同深度,多項(xiàng)式函數(shù)的MAE值均為最小,分別是0.26、0.23和0.25,并且在25 cm土壤深度時(shí)取得最小值。
綜合5、25和60 cm 3種土壤深度的RMSE和MAE值可以發(fā)現(xiàn),5種函數(shù)中,多項(xiàng)式函數(shù)的RMSE和MAE在5、25和60 cm 3種土壤深度下的值均最小,表明采用多項(xiàng)式函數(shù)擬合干邊和濕邊計(jì)算的特征空間的結(jié)果精度最高,并且在25 cm深度時(shí),RMSE和MAE值均為3組數(shù)據(jù)中的最小值,表明相對(duì)于5和60 cm土壤深度,特征空間計(jì)算的結(jié)果能夠較好地反演該研究區(qū)內(nèi)25 cm深度的土壤含水率。
根據(jù)上述結(jié)果分析可知,在采用線性、指數(shù)、對(duì)數(shù)、多項(xiàng)式和冪5種函數(shù)對(duì)干邊和濕邊進(jìn)行擬合的方式中,多項(xiàng)式函數(shù)的擬合效果最好,精度最高。根據(jù)表4中5種函數(shù)的RMSE和MAE值可以發(fā)現(xiàn),采用多項(xiàng)式函數(shù)擬合計(jì)算的土壤含水率的精度較線性函數(shù)的計(jì)算結(jié)果有一定的提升。對(duì)于5 cm土壤深度,多項(xiàng)式函數(shù)的RMSE和MAE分別為0.29、0.26,線性函數(shù)分別為0.31、0.29;對(duì)于25 cm土壤深度,多項(xiàng)式函數(shù)的RMSE和MAE分別為0.27、0.23,線性函數(shù)分別為0.29、0.26;對(duì)于60 cm土壤深度,多項(xiàng)式函數(shù)的RMSE和MAE分別為0.28、0.25,線性函數(shù)分別為0.30、0.28;通過(guò)對(duì)比可以發(fā)現(xiàn),對(duì)于3種不同土壤深度,多項(xiàng)式函數(shù)和線性函數(shù)之間RMSE和MAE差值較小,即誤差整體相差較小,精度略有提升。在本研究案例中,采用線性函數(shù)和多項(xiàng)式函數(shù)分別對(duì)LST/NDVI特征空間中的干邊和濕邊擬合的決定系數(shù)r2差異較小(表3)。其中,線性函數(shù)擬合干邊的r2為0.60,濕邊r2為0.63;多項(xiàng)式函數(shù)擬合干邊的r2為0.67,濕邊r2為0.70。但并非所有的研究區(qū)域和研究時(shí)段,多項(xiàng)式函數(shù)和線性函數(shù)之間的差異均較小,根據(jù)王行漢等[39]的研究,在對(duì)于中國(guó)南方地區(qū)的研究中,采用線性函數(shù)擬合方式獲取的干邊方程擬合r2為0.946 4,濕邊方程擬合r2為0.16;采用多項(xiàng)式函數(shù)擬合r2為0.998,濕邊方程的擬合r2為0.970 5,通過(guò)對(duì)比,對(duì)于干邊的擬合2種方法沒(méi)有特別明顯的差異r2基本處于0.9以上,但對(duì)于濕邊的擬合2種方式差異較大,對(duì)模型計(jì)算的結(jié)果有較大影響。
因此,可以發(fā)現(xiàn)采用線性函數(shù)擬合干邊和濕邊方程的精度在不同的研究區(qū)域和不同的研究時(shí)間上存在一定的不確定性,從而對(duì)模型的計(jì)算結(jié)果造成一定的偏差。傳統(tǒng)的模型構(gòu)建中,一般采用線性函數(shù)的擬合方式獲取干邊和濕邊方程,然而采用該方式趨勢(shì)線可靠性無(wú)法得到保證,如果數(shù)據(jù)自身線性趨勢(shì)性較好,擬合結(jié)果精度則較高;反之,如果數(shù)據(jù)自身線性趨勢(shì)性較差,擬合結(jié)果精度則較低,其計(jì)算結(jié)果受數(shù)據(jù)自身影響較大,從而為計(jì)算結(jié)果帶來(lái)一定的不確定性。
上述討論表明,本論文研究提出的基于多項(xiàng)式函數(shù)擬合干邊和濕邊方程的方法相對(duì)于傳統(tǒng)的線性函數(shù)的擬合方式穩(wěn)定性強(qiáng),過(guò)程關(guān)鍵步驟可控,對(duì)不同的研究區(qū)域和不同的研究時(shí)間均可保障模型計(jì)算結(jié)果的精確性,對(duì)提升該模型的適用性具有重要意義。
采用線性、多項(xiàng)式、對(duì)數(shù)、指數(shù)和冪函數(shù)5種不同函數(shù)分別對(duì)特征空間的干邊和濕邊進(jìn)行了擬合,并結(jié)合土壤墑情觀測(cè)站點(diǎn)數(shù)據(jù)對(duì)美國(guó)俄克拉荷馬州進(jìn)行了應(yīng)用,結(jié)果表明:
1)特征空間中構(gòu)建的干邊,即最大地表溫度組成的離散點(diǎn),傾向于多項(xiàng)式分布,表現(xiàn)在多項(xiàng)式擬合的決定系數(shù)r2在5種擬合方式中最高,達(dá)到0.67;特征空間中構(gòu)建的濕邊,即最小地表溫度組成的離散點(diǎn),傾向于多項(xiàng)式分布,表現(xiàn)在多項(xiàng)式擬合的r2在5種擬合方式中最高,達(dá)到0.70。
2)通過(guò)5種函數(shù)對(duì)干邊和濕邊進(jìn)行擬合,計(jì)算LST/NDVI特征空間值,并利用研究區(qū)內(nèi)的16個(gè)土壤墑情站點(diǎn)5、25和60 cm不同深度的3組數(shù)據(jù)分別與5種函數(shù)擬合計(jì)算的結(jié)果進(jìn)行對(duì)比,結(jié)果顯示多項(xiàng)式擬合構(gòu)建特征空間計(jì)算結(jié)果的均方根誤差和平均絕對(duì)誤差值均最小,表明采用多項(xiàng)式函數(shù)擬合干邊和濕邊計(jì)算的特征空間的結(jié)果精度最高,并且在25 cm深度時(shí),RMSE和MAE值均為3組數(shù)據(jù)中的最小值,表明相對(duì)于5和60 cm土壤深度,特征空間計(jì)算的結(jié)果能夠較好地反演該研究區(qū)內(nèi)25 cm深度的土壤含水率。
[1] Price J C. Using spatial context in satellite data to infer regional scale evapotranspiration [J]. IEEE Transactions on Geoscience and Remote Sensing, 1990, 28: 940-948.
[2] Carlson T N. An overview of the triangle method for estimating surface[J]. sensor, 2007, 7: 1612-1629.
[3] Goward S N, Cruickshanks G D, Hope A S. Observed relation between thermal emission and reflected spectral radiance of a complex vegetated landscape[J]. Remote Sensing of Environment, 1985, 18(2): 137-146.
[4] Nemani R R, Running S W. Estimation of regional surface resistance to evapotranspiration from NDVI and thermal-IR AVHRR data[J]. Journal of Applied meteorology, 1989, 28(4): 276-284.
[5] Goward S N, Hope A S. Evapotranspiration from combined reflected solar and emitted terrestrial radiation: Preliminary FIFE results from AVHRR data[J]. Advances in Space Research, 1989, 9(7): 239-249.
[6] Carlson T N, Perry E M, Schmugge T J. Remote estimation of soil moisture availability and fractional vegetation cover for agricultural fields[J]. Agricultural and Forest Meteorology, 1990, 52(1): 45-69.
[7] 田苗,王鵬新,孫威. 基于地表溫度與植被指數(shù)特征空間反演地表參數(shù)的研究進(jìn)展[J]. 地球科學(xué)進(jìn)展,2010,25(7):698-705. Tian Miao, Wang Pengxin, Sun Wei. A review of retrieving of land surface parameters using the land surface temperature-vegetation index feature space[J]. Advanceds in Earth Science, 2010, 25(7): 698-705. (in Chinese with English abstract)
[8] 楊曦,武建軍,閆峰,等. 基于地表溫度植被指數(shù)特征空間的區(qū)域土壤干濕狀況[J]. 生態(tài)學(xué)報(bào),2009,29(3):1205-1216. Yang Xi, Wu Jianjun, Yan Feng, et al. Assessment of regional soil moisture status based on characteristics of surface temperature/vegetation index space[J]. Acta Ecologica Sinica, 2009, 29(3): 1205-1216. (in Chinese with English abstract)
[9] Gillies R, Kustas W, Humes K. A verification of the'triangle'method for obtaining surface soil water content and energy fluxes from remote measurements of the normalized difference vegetation index (NDVI) and surface[J]. International Journal of Remote Sensing, 1997, 18(15): 3145-3166.
[10] Boegh E, Soegaard H, Hanan N, et al. A remote sensing study of the NDVI–Ts relationship and the transpiration from sparse vegetation in the Sahel based on high-resolution satellite data[J]. Remote sensing of Environment, 1999, 69(3): 224-240.
[11] Sandholt I, Rasmussen K, Andersen J. A simple interpretation of the surface temperature/vegetation index space for assessment of surface moisture status[J]. Remote Sensing of Environment, 2002, 79(2): 213-224.
[12] 于敏,高玉中,張洪玲. 地表溫度-植被指數(shù)特征空間干旱監(jiān)測(cè)方法的改進(jìn)[J]. 農(nóng)業(yè)工程學(xué)報(bào),2010,26(9):243-250. Yu Min, Gao Yuzhong, Zhang Hongling, et al. Improvement of Ts-NDVI space method in drought monitoring[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2010, 26(9): 243-250. (in Chinese with English abstract)
[13] 李紅軍,鄭力,雷玉平,等. 植被指數(shù)-地表溫度特征空間研究及其在旱情監(jiān)測(cè)中的應(yīng)用[J]. 農(nóng)業(yè)工程學(xué)報(bào),2006,22(11):170-174. Li Hongjun, Zheng Li, Lei Yuping, et al. Vegetation index-surface temperature feature space and its application in the regional drought monitoring[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2006, 22(11): 170-174. (in Chinese with English abstract)
[14] Prihodko L, Goward S N. Estimation of air temperature from remotely sensed surface observations[J]. Remote Sensing of Environment, 1997, 60(3): 335-346.
[15] Otkin J A, Anderson M C, Hain C, et al. Assessing the evolution of soil moisture and vegetation conditions during the 2012 United States flash drought[J]. Agricultural and Forest Meteorology, 2016(218/219): 230-242.
[16] Stisen S, Sandholt I, Nrgaard A, et al. Combining the triangle method with thermal inertia to estimate regional evapotranspiration: Applied to MSG-SEVIRI data in the senegal river basin[J]. Remote Sensing of Environment, 2008, 112(3): 1242-1255.
[17] Minacapilli M, Consoli S, Vanella D, et al. A time domain triangle method approach to estimate actual evapotranspiration:Application in a mediterranean region using MODIS and MSG-SEVIRI products[J]. Remote Sensing of Environment, 2016, 174: 10-23.
[18] 王飛,丁建麗,伍漫春. 基于NDVI-SI 特征空間的土壤鹽漬化遙感模型[J]. 農(nóng)業(yè)工程學(xué)報(bào),2010,26(8):168-173. Wang Fei, Ding Jianli, Wu Manchun. Remote sensing monitoring models of soil salinization based on NDVI-SI feature space [J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2010, 26(8): 168-173. (in Chinese with English abstract)
[19] 劉云,孫丹峰,宇振榮,等. 基于NDVI-Ts特征空間的冬小麥水分診斷與長(zhǎng)勢(shì)監(jiān)測(cè)[J]. 農(nóng)業(yè)工程學(xué)報(bào),2008,24(5):147-151. Liu Yun, Sun Danfeng, Yu Zhenrong, et al. Water deficit diagnosis and growing condition monitoring of winter wheat based on NDVI-Ts feature space[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2008, 24(5): 147-151. (in Chinese with English abstract)
[20] Zhao J, Xu J, Xie X, et al. Drought monitoring based on TIGGE and distributed hydrological model in Huaihe River Basin, China [J]. Sci Total Environ, 2016, 553: 358-365.
[21] Zhang T, Lin X. Assessing future drought impacts on yields based on historical irrigation reaction to drought for four major crops in Kansas[J]. Sci Total Environ, 2016, 550: 851-860.
[22] Mart nez-fernndez J, Gonz lez-zamora A, Snchez N, et al. Satellite soil moisture for agricultural drought monitoring: Assessment of the SMOS derived soil water deficit index[J]. Remote Sensing of Environment, 2016, 177: 277-286.
[23] Assal T J, Anderson P J, Sibold J. Spatial and temporal trends of drought effects in a heterogeneous semi-arid forest ecosystem [J]. Forest Ecology and Management, 2016, 365: 137-151.
[24] Carlson T N, Gillies R R, Schmugge T J. An interpretation of methodologies for indirect measurement of soil water content[J]. Agricultural and Forest Meteorology, 1995, 77: 191-205.
[25] Jiang L, Islam S. A methodology for estimation of surface evapotranspiration over large areas using remote sensing observations[J]. Geophysical Research Letters, 1999, 26: 2773-2776.
[26] Jiang L, Islam S, Guo W, et al. A satellite-based daily actual evapotranspiration estimation algorithm over South Florida[J]. Global and Planetary Change, 2009, 67(1): 62-77.
[27] Long D, Singh V P. A modified surface energy balance algorithm for land (M-SEBAL) based on a trapezoidal framework [J]. Water Resources Research, 2012, 48(2): doi:10.1029/2011WR010607.
[28] Moran M, Clarke T, Inoue Y, et al. Estimating crop water deficit using the relation between surface-air temperature and spectral vegetation index[J]. Remote Sensing of Environment, 1994, 49(3): 246-263.
[29] Arnon K, Nurit A, Rachelt P, et al. Use of NDVI and land surface temperature for drought assessment: Merits and limitations[J].Journal of Climate, 2010, 23(3): 618-633.
[30] Liu Y, Mu X, Wang H, et al. A novel method for extracting green fractional vegetation cover from digital images[J]. Journal of Vegetation Science, 2012, 23(3): 406-418.
[31] Zhang X, Liao C, Li J, et al. Fractional vegetation cover estimation in arid and semi-arid environments using HJ-1 satellite hyperspectral data[J]. International Journal of Applied Earth Observation and Geoinformation, 2013, 21: 506-512.
[32] 白潔,劉紹民,扈光. 針對(duì) TM/ETM+ 遙感數(shù)據(jù)的地表溫度反演與驗(yàn)證[J]. 農(nóng)業(yè)工程學(xué)報(bào),2008,24(9):148-154. Bai Jie, Liu Shaomin, Hu Guang. Inversion and verification of land surface temperature with remote sensing TM/ETM+ data[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2008, 24(9): 148-154. (in Chinese with English abstract)
[33] Sobrino J A, Jim nez-mu oz J C, Paolini L. Land surface temperature retrieval from landsat tm 5[J]. Remote Sensing of Environment, 2004, 90(4): 434-440.
[34] Liu Y, Mu X, Wang H, et al. A novel method for extracting green fractional vegetation cover from digital images[J]. Journal of Vegetation Science, 2012, 23(3): 406-418.
[35] 穆少杰,李建龍,陳奕兆,等. 2001-2010 年內(nèi)蒙古植被覆蓋度時(shí)空變化特征[J]. 地理學(xué)報(bào),2012,67(9):1255-1268. Mu Shaojie, Li Jianlong, Chen Yizhao, et al. Spatial differences of variations of vegetation coverage in Inner Mongolia during 2001-2010[J]. Acta Geographica Sinica, 2012, 67(9): 1255-1268. (in Chinese with English abstract)
[36] Dennison P E, Roberts D A. Endmember selection for multiple endmember spectral mixture analysis using endmember average RMSE[J]. Remote Sensing of Environment, 2003, 87(2): 123-135.
[37] 王行漢,叢沛桐,劉超群,等. 2004-2013年珠江流域植被變化及其脅迫分析[J]. 生態(tài)學(xué)報(bào),2017,37(19),doi:10.5846/stxb201607111416. Wang Xinghan, Cong Peitong, Liu Chaoqun, et al. Analysis of vegetation variation and stress factors in the pearl river basin from 2004 to 2013[J]. Acta Ecologica Sinica, 2017, 37(19), doi:10.5846/stxb201607111416.
[38] Liu D, Mishra A K, Yu Z. Evaluating uncertainties in multi-layer soil moisture estimation with support vector machines and ensemble Kalman filtering[J]. Journal of Hydrology, 2016, 538: 243-255.
[39] 王行漢,亢慶,扶卿華,等. 一種農(nóng)業(yè)旱情遙感監(jiān)測(cè)方法:CN105929406A[P]. 2016-09-07.
Discussion on method of nonlinear fitting dry and wet edges of LST/ NDVI feature space better than traditional linear fitting method
Wang Xinghan1,2,3, Cong Peitong1※, Kang Qing2, Fu Qinghua2, Liu Chaoqun2, Wang Xiaogang2
(1. College of Water Conservancy and Civil Engineering, South China Agriculture University, Guangzhou 510610, China; 2. Pearl River Institute of Hydraulic Research, Pearl River Water Resources Commission,Ministry of Water Resources, Guangzhou 510611, China; 3. Key Laboratory of the Pearl River Estuarine Dynamics and Associated Process Regulation, Ministry of Water Resources, Guangzhou 510611, China)
Land surface temperature / vegetation index feature space has important applications in quantitative retrieval of water content in soil and crop evapotranspiration. However, at present, the research on the fitting of the dry and wet edges of the land surface temperature/vegetation index feature space was relatively lacking. In the tradition, for dry edge of the model, a simple linear negative correlation was adopted to fit the parameters, and wet edge was considered as a simplified treatment parallel to the coordinate axis. Whether it is appropriate is the focus of this paper that needs to be discussed. The study area was located in Oklahoma, the United States. Based on Landsat TM5 image data, land surface temperature (LST) and normalized difference vegetation index (NDVI) were calculated, and LST was calculated by radiation equation model and NDVI by pixel dichotomy model. And the fitting of dry edges and wet edges of LST/NDVI feature space was carried out with different functions, which included linear function, exponential function, logarithm function, power function and polynomial function. All of them were used to fit dry edges and wet edges respectively, and the results were evaluated by the measured data of water content in soil. The results showed that for the fitting of 5 different functions, r2value as a whole was between 0.4 and 0.7, and there were some differences in the fitting precision between different fitting methods. For the fitting of dry edges, r2value of exponential function fitting was 0.64, r2value of linear function fitting was 0.60, r2value of logarithm function fitting was 0.41, r2value of power function fitting was 0.43, and r2value of polynomial function fitting was 0.67 which was the best fitting way for dry edges. For the fitting of wet edges, r2value of exponential function fitting was 0.59, r2value of linear function fitting was 0.63, r2value of logarithm function fitting was 0.67, r2value of power function fitting was 0.69, and r2value of polynomial function fitting was 0.70 which was the best fitting way for wet edges. For the fitting of dry edges and wet edges, polynomial function was the best method. And the results of 5 kinds of function fitting were compared with those from the soil moisture stations in the study area. Root mean square error (RMSE) and mean absolute error (MAE) were calculated, and 5, 25 and 60 cm depth were selected. In the 3 different depths, RMSE and MAE of polynomial function were the smallest. RMSE at 5 cm depth was 0.29, RMSE at 25 cm depth was 0.27, and RMSE at 60 cm depth was 0.28; MAE at 5 cm depth was 0.26, MAE at 25 cm depth was 0.23, and MAE at 60 cm depth was 0.25. The results indicated that the LST/NDVI feature space inversion based on dry edges and wet edges fitting with the polynomial function was the most accurate for the soil surface water content in this study area, and it was most sensitive to water content at 25 cm depth in soil. For an optimal fitting, it must be an optimal solution between fitting accuracy and fitting efficiency. In the process of this study, only small amount of data were involved, so the main consideration was the accuracy of dry edges and wet edges fitting, not taking into account the time cost of computer computing process. But for the large amount of data operations in the actual application process, the time efficiency still needs to be considered.
vegetation; temperature; soil moisture; land surface temperature/vegetation index; feature space; dry edge; wet edge
10.11975/j.issn.1002-6819.2017.11.039
TP79
A
1002-6819(2017)-11-0306-09
2016-10-02
2017-05-04
廣東省水利科技創(chuàng)新項(xiàng)目(2016-09);廣州市科技計(jì)劃項(xiàng)目(201605030009)
王行漢,男,江蘇東臺(tái)人,工程師,博士生,主要從事植被遙感、水利遙感和農(nóng)業(yè)旱情遙感監(jiān)測(cè)。廣州 華南農(nóng)業(yè)大學(xué)水利與土木工程學(xué)院,510611。Email:rsgiswxh@126.com
※通信作者:叢沛桐,教授,博士生導(dǎo)師,主要研究方向?yàn)樗畔⒒c防災(zāi)減災(zāi)。廣州 華南農(nóng)業(yè)大學(xué)水利與土木工程學(xué)院,510610。
Email:congpeitong@126.com